Automated Segmentation and Quantification of Epithelial Cell Nuclei in Images of Pap Smears Stained with Immunoreactive Biomarkers

Tech ID: ger000585

 

Introduction

This invention addresses the problem of segmentation of nuclei and quantification of respective nuclear signals in digitized preparations of human specimens imaged in two dimensions (2-D).

 

The invention is an image analysis method utilizing gradient-based radial symmetry followed by adaptive processing to first automatically outline cell nuclei and then quantify distribution of biomarkers within the outlined nuclear areas for the assessment of cellular immunoreactivity in liquid-based Pap tests.

 

Technology Description

The nuclei can be distinguished from cytoplasm by their high circularity and low optical density features. When applied together, these two can provide a more robust detection than techniques that rely on image intensity only and do not consider the target shape of the object (nucleus) being detected. This method is primarily intended to assess cervical cancer risk as well as for rapid and unsupervised screening of cells for basic research, clinical diagnostics and prognostic tasks.

Currently researchers have tested and evaluated this method on Pap tests. Its purpose is to: (i) improve the sensitivity and specificity of precancerous and cancerous cell detection, (ii) reduce burden of manual slide screening, and (iii) increase robustness and throughput of automated Pap test screening procedures.

 

Advantages/Novelty

• This method performs more reliably than existing methods for cells with little to no overlap, whose performance may degrade for clustered cells, or closely positioned nuclei.

• Existing attempts to the automated outline of nuclei in Pap smear images involve algorithms that solely utilize locally decreased image intensity attributed to the presence of a nucleus as a key image feature to be detected. Current techniques are based on k-means clustering, fuzzy logic, support vector machines, region growing, local image thresholding and active contours applied to either conventional (color or monochromatic), as well as spectral images of cells.

 

Applications

• The algorithm could be built into existing automated Pap test slide screening devices.

• Faster and more reliable screening can reduce dollar amount paid per analysis of one slide.

 

Intellectual Property 

Registered copyright (TXu 1-809-420).

 

Publications 

Qin et al. Computerized Delineation of Nuclei in Liquid-Based Pap Smears Stained with Immunohistochemical Biomarkers. Cytometry B Clin Cytom. 2015, 88(2):110-9. [LINK]

Gertych et al. Automated detection of dual p16/Ki67 nuclear immunoreactivity in liquid-based Pap tests for improved cervical cancer risk stratification. Ann Biomed Eng. 2012, 40(5):1192-204. [LINK]

 

 

 

Patent Information:
For Information, Contact:
Wenyue Du
Senior Associate - IP Management & Licensing
310-423-2241
wenyue.du@cshs.org
Inventors:
Arkadiusz Gertych
Keywords:
Imaging
Oncology
xCytopathology
xPathology
xTwo-Dimensional Cell or Tissue Imaging